Homomorphic Encryption in the Open Source World
summary
A Red Hat research effort compares open libraries and builds proofs of concept: Microsoft SEAL is the C++ leader, older Paillier-style code is ~1200× slower, and Python-friendly options include Pyfhel, Pyfhe and especially TenSEAL which runs CKKS on vectors and enables encrypted logistic regression. Encrypted evaluation shows a 0.27 accuracy gap versus plaintext, training epochs can run ~350s per epoch on a small dataset, and core limits remain: no reliable encrypted comparison or cheap division.
AI-written summary · about
chapters (10)
- 0:00Introduction to HE
- 1:20What is Homomorphic Encryption
- 2:54Applications and Healthcare
- 4:15HE in the Open Source World
- 7:10Limitations of the Technology
- 8:23Types of HE Schemes
- 12:23Open Source HE Libraries
- 17:32Code Demonstration and Results
- 24:55HE vs Federated Learning
- 28:15Q&A and Conclusion
related talks
37mFrontiers in Cryptography
37mOpen-source Machine Learning on Encrypted Data
40mConfidential Computing and Privacy-Enhancing Technologies - The Landscape
41mNon-traditional Encryption Problems And Solutions
18mSimEnc: A High-Performance Similarity-Preserving Encryption Approach for...
44mIntroduction to confidential computing – protecting sensitive workloads
35mEncrypted Computation: What if Decryption Wasn’t Needed?
40mTrust No One